Aditya Sriram recently shared something that caught my attention: "This Meta Ad Tracker system is absolutely insane!" He went on to say it "scrapes your competitors' video ads from the Facebook Ad Library... And runs AI analysis on each one automatically." If you have ever done competitor research the old way, you know exactly why that line resonates.

"You’re manually clicking through the Ad Library. Watching competitor videos one by one.
Trying to figure out which hooks they’re testing... By the time you’ve analyzed 10 ads, you’re exhausted."

That is the real problem: not the lack of data, but the friction of turning a huge pile of creatives into usable decisions. Aditya’s point is simple and practical: if research is too slow and too manual, you do less of it, you do it inconsistently, and your creative testing suffers.

In this post, I want to expand on what Aditya described and translate it into a blog-friendly playbook: what an AI-powered Meta Ad Library tracker actually changes, what it can (and cannot) tell you, and how to use outputs like hooks, visual notes, and CTAs to build better tests.

The bottleneck in competitor research is not access, it is analysis

The Facebook Ad Library is a gift and a trap.

It is a gift because you can see real ads from real brands in-market right now.
It is a trap because the interface is built for browsing, not for systematic learning.

Aditya described the typical workflow: click, watch, guess the hook, take notes, repeat. After 10 ads, your brain is done, and the notes are scattered across docs that never become a testing plan.

What marketers actually need is a repeatable pipeline:

Collect creatives reliably (and frequently)
Summarize them consistently (same categories every time)
Compare patterns across many ads
Turn patterns into testable hypotheses

AI analysis is most valuable when it compresses steps 2 and 3, so you can spend your time on step 4.

What Aditya’s system is doing (in plain English)

Aditya outlined a straightforward flow:

Ask it to analyze competitor ads
Choose a lookback window (7 days, 30 days, or all time)
Scrape video ads, thumbnails, and ad text
Then Gemini Vision "watches" each ad and breaks it down

The important shift is that the system treats ads like structured data, not like a playlist you have to binge.

Instead of "I watched 10 ads," you get: "I analyzed 150 ads and extracted the same fields for each." That is the difference between anecdotal inspiration and actual competitive intelligence.

The five creative fields that matter (and why they matter)

Aditya listed what the system pulls and analyzes:

Hook excerpt (what grabbed attention)
Visual notes (what is happening on screen)
Primary message and target audience
Creative approach and key elements
CTA and emotional tone

Let’s unpack how each field becomes actionable.

1) Hook excerpt: your scroll-stopping inventory

Hooks are not just clever lines. They are patterns of attention.

When you have hook excerpts across dozens of ads, you can cluster them into categories, for example:

"Problem first" ("Still struggling with...")
"Outcome first" ("Get X in 7 days")
"Contrarian" ("Stop doing X")
"Social proof" ("100,000 customers")
"Mechanism" ("We use Y to do Z")

Your test plan becomes clearer: you are not testing random ideas, you are testing hook families.

2) Visual notes: what is doing the persuading

Most teams over-index on copy and under-index on what is actually on screen. Visual notes force you to notice patterns like:

UGC face-to-camera vs polished studio
Fast cuts vs slow demo
Text overlays density
Before/after sequences
Product-in-hand demonstrations

If AI can consistently describe these elements, you can build a creative brief that mirrors what is working in the category, without copying.

3) Primary message and audience: what the ad is really selling

Two brands can advertise the same product, but sell different beliefs.

AI summaries help you answer:

Is the core promise speed, price, status, safety, simplicity, or performance?
Who is the ad speaking to: beginners, experts, bargain hunters, busy parents, founders?

This matters because your creative will fail if you borrow the wrong promise for your audience.

4) Creative approach and key elements: the repeatable template

This is where you find the "format" that keeps showing up, such as:

Hook -> proof -> demo -> offer
Founder story -> problem -> solution -> credibility
Meme-style pattern interrupt -> quick benefit stack -> CTA

Once you spot a template, you can produce variations faster: swap hook, swap proof type, swap demo angle, while keeping the skeleton intact.

5) CTA and emotional tone: what pushes the click

Aditya called out CTA and emotional tone, and that is a subtle but important point.

Many ads are not aggressive. They are calm, reassuring, and friction-reducing. Others are urgent and challenge-based.

If your brand voice is misaligned with the dominant emotional tone in the category, your ads can feel "off" even when the offer is good.

A practical way to use an AI ad tracker (without fooling yourself)

AI makes research faster, but you still need a method. Here is a workflow I would recommend if you had the kind of system Aditya described.

1) Set a weekly cadence and a focused competitor list

Pick 5 to 15 advertisers that match:

Your price point and positioning
Your customer sophistication level
Your funnel type (lead gen vs ecommerce)

Then run a 7-day lookback weekly. The goal is to catch new tests, not to build a museum of old ads.

2) Turn outputs into a "testing backlog"

For each ad summary, capture:

Hook category
Creative format
Offer angle
Proof type (testimonial, data, founder, demo)

Then convert patterns into hypotheses like:

"In this niche, outcome-first hooks paired with quick demos appear most frequently. We will test 5 variations with our product demo and 3 different outcome claims."

3) Use the system for clustering, not final answers

AI is good at labeling and grouping. It is not always good at being right.

Use it to:

Extract consistent fields
Identify recurring patterns
Flag ads you should watch personally

Then do a human pass on the top 10 percent of ads that look most relevant. Aditya’s system reduces the haystack so you can inspect the needles.

4) Create a creative brief from competitive signals

Instead of sending your team "cool ads," send a brief:

3 hook families to test
2 dominant visual styles to try
1 or 2 emotional tones to match or intentionally contrast
A list of must-include proof elements

That is how competitor research becomes production velocity.

What this approach does not solve (and what to watch out for)

Even with scraping and AI vision analysis, there are limitations:

Ad Library does not reveal full targeting, budgets, or true performance.
You can misread what is working if you assume "more ads" equals "winning ads." Some are just spend experiments.
AI can hallucinate details (especially nuanced claims or fast-moving on-screen text). Always validate key takeaways by watching the original.
Scraping and automation should respect platform terms and privacy boundaries. Build responsibly.

The best mindset is: treat AI summaries as research assistants, not judges.

The bigger takeaway from Aditya’s post

Aditya’s viral post was not just about a cool tool. It was about changing the unit of work.

Instead of "I watched some competitor ads," the new unit becomes "I have a structured dataset of creative angles and patterns, refreshed weekly." That shift is what makes creative strategy feel less like guesswork and more like a system.

If you are running paid social in 2026, your advantage is not seeing ads. Everyone can see ads. Your advantage is learning faster than everyone else, and turning learning into better creative tests.

This blog post expands on a viral LinkedIn post by Aditya Sriram, Building GoMarble || AI Agent for paid media marketers; built on your Meta Ads, Google Ads, Shopify, and GA4.. View the original LinkedIn post →